Glycophorin A transmembrane domain dimerization in plasma membrane vesicles derived from CHO, HEK 293T, and A431 cells

Abstract

Membrane protein interactions, which underlie biological function, take place in the complex cellular membrane environment. Plasma membrane derived vesicles are a model system which allows the interactions between membrane proteins to be studied without the need for their extraction, purification, and reconstitution into lipid bilayers. Plasma membrane vesicles can be produced from different cell lines and by different methods, providing a rich variety of native-like model systems. With these choices, however, questions arise as to how the different types of vesicle preparations affect the interactions between membrane proteins. Here we address this question using the glycophorin A transmembrane domain (GpA) as a model system. We compare the dimerization of GpA in six different vesicle preparations derived from CHO, HEK 293T and A431 cells. We accomplish this with the use of a FRET-based method which yields the FRET efficiency, the donor concentration, and the acceptor concentration in each vesicle. We show that the vesicle preparation protocol has no statistically significant effect on GpA dimerization. Based on these results, we propose that any of six the plasma membrane preparations investigated here can be used as a model system for studies of membrane protein interactions.

Introduction

Membrane proteins (MPs) fold in a complex heterogeneous environment (1, 2). A useful paradigm for membrane protein folding is the two-stage model which assumes that folding occurs via (i) insertion of the hydrophobic transmembrane (TM) helices in membranes, followed by (ii) lateral association of the inserted helices into functional membrane proteins (3). There have been numerous studies of TM helix lateral interactions, which have contributed significantly towards the understanding of membrane protein folding (4–8). Such TM helix dimerization studies have been performed in a variety of hydrophobic environments including detergents, bicelles, lipid bilayers, and bacterial membranes (9–14). In our lab, we have performed measurements in mammalian membranes using a FRET-based technique called Quantitative Imaging FRET (QI-FRET) (15, 16). In these experiments, cells are transfected with genes encoding for TM helices tagged with fluorescent proteins. After the proteins are trafficked to the plasma membrane, the cells are incubating in a vesiculation buffer. As a result, the cells shed vesicles as part of their apoptotic response to the vesiculation buffer stress (17–19). The membranes of these vesicles contain various membrane proteins and mimic the natural crowded membrane environment. We have demonstrated that these vesicles can serve as a plasma membrane model to study the thermodynamics of TM helix dimerization, as the distribution of the proteins in the vesicles is homogeneous and the data are well described by models assuming monomer-dimer equilibrium in the membrane (15, 16).

The most widely used vesiculation protocol uses DTT and formaldehyde (17). An alternative osmotic vesiculation buffer has been described in the literature by Cohen and co-workers, but it works only for A431 cells (20). We have recently developed a new vesiculation buffer, containing chloride salts, which can be used for both A431 and CHO cells (21). Thus, plasma membrane derived vesicles can be now produced from different cell lines and with different methods, providing a variety of native-like model systems. Questions therefore arise (1) whether MP interactions differ in different preparations of plasma derived vesicles and (2) whether a particular vesicle preparation should be preferred over others for studies of MP interactions.

Many of the principles behind TM helix dimerization have emerged from studies of the TM domain of the erythrocyte glycoprotein glycophorin A (GpA) (9, 10). Interestingly, GpA dimerization has been shown to be very sensitive to the lipid composition. Effects as large as 9 kcal/mole were measured upon the addition of charged lipids, non-bilayer lipids, or bacterial membrane proteins (13). In this study we asked whether the cell line used to produce plasma membrane derived vesicles, or the particular method of vesiculation, affects the dimerization of GpA. To answer this question, we measured and compared the dimerization of GpA in six different vesicle preparations derived from CHO, HEK 293T and A431 cells.

Materials and Methods

Plasmid constructs

The GpA plasmid was created as described previously(15) The eYFP plasmid was a kind gift from Dr. M. Betenbaugh (Johns Hopkins University, Baltimore, MD) and the pRSET B-mCherry plasmid was a gift from Dr. R. Tsien (University of California, San Diego). All of the plasmids used for mammalian expression were constructed with the pcDNA 3.1(+) vector (Invitrogen). All primers were purchased from Invitrogen.

The cloning procedure used to create the plasmids pcDNA-GpA-eYFP and pcDNA-GpA-mCherry is published (15). The plasmids encoded for a N-terminal signal peptide directing GpA to the plasma membrane, the GpA TM domain (sequence: LIIFGVMAGVIGTILLISYGIRRL), a flexible 15 amino acid-long (GGS)₅ linker, and a fluorescent protein (either eYFP and mCherry). The bacterial expression and purification of the soluble eYFP and mCherry used for calibration of concentrations was performed as described in (22).

Cell growth and transfection

Chinese Hamster Ovary (CHO), Human Embryonic Kidney 293T (HEK 293T) and A431 cells were stored in liquid nitrogen and thawed when needed. The cells were cultured and grown to confluency, and passed five times prior to the experiments. After 35 passages, the cells were discarded.

Cells were grown at 37 °C with 5% CO₂. 4 × 10⁵ cells were seeded per well in a 6-well plate 24h before transfection. Transfection was carried out using Fugene HD transfection reagent (Promega.Corporation), following the manufacturer’s protocol. Cells were co-transfected with pcDNA-GpA-eYFP and pcDNA-GpA-mCherry, and vesiculated 24 h post transfection. HEK293T and CHO cells were transfected with a total of 3 ug of the plasmid DNA contsructs while A431 cells were transfected with a total of 6 ug of DNA.

Vesiculation

Vesiculation was performed using three different methods (17, 20, 21). To vesiculate with the DTT/formaldehyde method developed by Scott (17–19), 70% confluent HEK293T, CHO and A431 cells were rinsed with PBS (pH 7.4) containing 0.75 mM calcium chloride and 0.5 mM magnesium chloride (CM-PBS), and incubated with 1 mL of a vesiculation buffer containing 25 mM formaldehyde, 0.5 mM 1,4-dithiotreitol (DTT) and protease inhibitor cocktail (Roche Applied Science) at 37 °C. To quench the formaldehyde after vesiculation, glycine solution in PBS was added to the vesiculation buffer to a final concentration of 0.125 M. A large number of vesicles were produced after 2 h, and the vesicles were transferred into 4-well Nunc Lab-Tek II chambered coverslips for imaging.

To vesiculate with the salt cocktail osmotic stress method of Cohen et al. (20), 70% confluent A431 cells were rinsed once with PBS (pH 7.4), and incubated with 1 mL of 10% PBS buffer at room temperature for 10 minutes. Subsequently the cells were incubated with 1 mL of hypertonic buffer containing 100 mM NaCl, 50 mM Na₂HPO₄, 0.5 mM KCl, 0.5 mM MgSO₄ and protease inhibitor cocktail adjusted to pH of 8.5. Vesicles were produced after 2 hours and imaged in 4-well chambered coverslips.

To vesiculate using the chloride salt osmotic stress method that we recently developed (21), 70% confluent CHO and A431 cells were rinsed three times with 30% PBS buffer at room temperature. The cells were then incubated with 1 mL of hypertonic buffer containing 200 mM NaCl, 5 mM KCl, 0.5 mM MgSO₄, 0.75 mM CaCl₂, 100 mM bicine and protease inhibitor cocktail adjusted to pH of 8.5. Vesicles were collected after 12h for imaging.

FRET measurements of dimerization propensities

Vesicles were imaged using a Nikon Eclipse confocal laser scanning microscope using a 60× water immersion objective. All the images were collected and stored at a 512 × 512 resolution. Three distinct scans were performed for each vesicle: (1) excitation at 488 nm, with a 500–530 nm emission filter (donor scan); (2) excitation at 488 nm, with a 565–615 nm emission filter (FRET scan); and (3) excitation at 543 nm, with a 650 nm longpass filter (acceptor scan) as discussed previously (16, 22).

The three images for each vesicle were analyzed using a Matlab® program to obtain the fluorescence intensity across the vesicle membrane (16, 22, 23). The fluorescence intensity across the membrane was modeled with a Gaussian function and the background intensity with as an error function. The integration of the Gaussians from the donor, acceptor, and FRET scans yielded the three fluorescence intensities, I_A, I_D, and I_FRET, for each vesicle. These three intensities were then used to calculate the donor and acceptor concentration in the vesicle as described in detail elsewhere (15).

Briefly, the concentration of the donor-labeled GpA per unit area in each vesicle was calculated according to:

$$C_D=\frac{I_D+G_F(I_{\mathit{FRET}}-{\beta }_D{I}_D-{\beta }_A{I}_A)}{i_D}$$ (1)

and the concentration of the acceptor-labeled GpA per unit area was determined as:

$$C_A=\frac{I_A}{i_A}$$ (2)

The coefficients i_D and i_A in equations (1) and (2) are calibration constants for the donor and the acceptor, respectively, and β_D and β_A are the donor and acceptor bleed-through coefficients. These parameters were determined by analyzing purified protein solutions of known concentrations (22). G_F is the gauge factor, a parameter which relates the sensitized emission of the acceptor to the quenching of the donor. It was calculated by analyzing vesicles loaded with a linked eYFP- mCherry soluble protein as described elsewhere (22).

The FRET efficiency for each vesicle is determined as:

$$E=1-\frac{I_D}{I_D+G_F(I_{\mathit{FRET}}-{\beta }_D{I}_D-{\beta }_A{I}_A)}$$ (3)

This FRET efficiency was then corrected for a FRET contribution which arises due to random proximity (within distances of 100 Å) of donor and acceptors. The proximity correction is based on a model by Wolber and Hudson (24) and has been experimentally verified (25). This correction yielded FRET due to sequence-specific dimerization, E_D. The dimeric fraction (i.e. the fraction of proteins that are dimeric) in each vesicle was then calculated as (26, 26, 27):

$$f=\frac{E_D}{x_A\overline{E}}.$$ (4)

Here x_A is the fraction of proteins labeled with the acceptor fluorophore, and Ē is the FRET efficiency in a dimer containing a donor and an acceptor. The value of Ē in equation (4) is known for the GpA construct used here, Ē = 0.63 ± 0.04, corresponding to a 48.5 Å separation distance between the fluorophores in the dimer (15).

The lateral dimerization of GpA in the membrane is described by the following reaction scheme:

$$M+M\stackrel{K_D}{\to }D,$$ (5)

where the dimerization constant is:

$$K_D=[D]/{[M]}^2,$$ (6)

and the total protein concentration is given by:

$$[T]=[M]+2[D].$$ (7)

K_D is determined by fitting the dimerization model given by equations (5) through (7) to the measured dimeric fractions (given by equation (4)). Defining the standard state as K_st = 1 nm²/protein, the free energy of dimer formation is calculated according to:

$$\mathrm{\Delta }G=-\mathit{RT}ln(\frac{K_D}{K_{st}})$$ (8)

Statistical Analysis

The errors in the calculated dimerization free energies were determined using Matlab, by (i) calculating the confidence interval for K_D when fitting the model to all the experimental dimeric fractions prior to binning and (ii) taking into account the experimental uncertainties in the fluorescence calibration parameters. The free energies calculated for all membrane preparations were compared using ANOVA. The parameters inputted into the ANOVA calculation were the free energies, the standard errors, and the number of data points (number of vesicles analyzed) which exceeded 200 in each case.

Results

Here we worked with six different plasma membrane vesicle preparations. Three of them utilized the DTT/formaldehyde vesiculation method and were produced from CHO, HEK293T, and A431 cells. The first cell line, CHO, is derived from the ovaries of Chinese Hamster and is perhaps the most widely used mammalian cell line for recombinant protein expression, as CHO cells grow fast and produce large protein quantities. The second cell line, HEK 293T, is a human cell line of kidney epithelial origin that is used for the transient and stable expression of human proteins. The third cell line, A431, is an epidermoid carcinoma cell line that is highly sensitive to mitogenic stimuli, and is usually avoided in studies of recombinant proteins. It was used here because this is the only cell line that can be vesiculated with the osmotic buffer of Cohen at al (20). Noteworthy, A431 cells express very high levels (3 × 10⁶ copies per cell) of the Epidermal Growth Factor Receptor (EGFR). This is a receptor which functions via lateral dimerization in the plasma membrane, and the homodimerization of its TM domains is believed to occurs through the GxxxG motif (28, 29), the very same sequence motif that drives GpA dimerization. Thus this cell line gave us the opportunity to access the likelihood of any “promiscuous” interactions between GpA and EGFR.

A431 cells were vesiculated using the three vesiculation methods described in Materials and Methods, while CHO cells were vesiculated using the DTT/formaldehyde and the chloride salt method. HEK 293T cells, on the other hand, could be vesiculated with the DTT/formaldehyde method only, as the cells quickly detached in the other two vesiculation buffers and thus the vesicles could not be isolated for imaging.

We measured the energetics of Glycophorin A TM domain dimerization in plasma membrane vesicles in the six different vesicle preparations discussed above. Cells were co-transfected with 3–6 ug DNA encoding GpA-TM-eYFP and GpA-TM-mCherry. After GpA expression and its trafficking to the plasma membranes of the cells, the cells were vesiculated as described in Materials and Methods. After vesicle production, single vesicles were imaged in a laser-scanning confocal fluorescence microscope such that three different images, donor, FRET, and acceptor images were captured for each vesicle as described (20). The donor concentration, the acceptor concentration, and the FRET efficiencies in each vesicle were determined using equations (1), (2), and (3) as outlined in Materials and Methods.

The measured FRET efficiencies are shown in Figure 1 as a function of the acceptor concentration. Each data point in Figure 1 corresponds to a single vesicle. Figure 1A compares the FRET efficiencies measured in vesicles derived from CHO, HEK 293T, and A431 cells upon incubation with the DTT/formaldehyde buffer. Figure 1B shows the FRET efficiencies in CHO cells vesiculated with the DTT/formaldehyde and the chloride salt osmotic stress buffer. Figure 1C compares the FRET efficiencies measured in A431 vesicles, produced with the three vesiculation buffers. These data are replotted in Supplemental Information.

Figure 1.

FRET efficiencies measured for GpA in plasma membrane derived vesicles, as a function of acceptor concentration. Genes encoding GpA attached to either YFP or mCherry at its C-terminus via a (GGS)₅ linker were expressed in cells. After GpA was trafficked to the plasma membrane, the cells were vesiculated using three different methods. (A) CHO, HEK293T and A431 vesicles, produced with the DTT/formaldehyde method of Scott (17). (B) CHO vesicles, produced with the DTT/formaldehyde method and the new chloride salt osmotic method that we have developed (21). (C) A431 cells vesiculated using the DTT/formaldehyde method, the chloride salt method, and the salt cocktail method developed by Cohen at al (20). Each data point corresponds to a single vesicle.

The calculation of dimeric fractions (i.e. the fraction of proteins that are dimeric) in each vesicle has been discussed in detail in several of our publications (16, 22, 23). The calculations follow the protocol that was originally developed to calculate free energies in lipid vesicles (27, 30). Briefly, the dimeric fraction was calculated for each vesicle using equation (4). Then, the calculated dimeric fractions were averaged within bins of total protein concentration of width = 5×10⁻⁴ molecules/nm². The averaged dimeric fractions as a function of total concentration, along with the standard errors, are shown in Figures 2A, 2B and 2C (and replotted in Supplemental Information).

Figure 2.

Dimeric GpA fractions as a function of total GpA concentration. The dimeric fractions were averaged in bins of 5 × 10 ⁻⁴ receptors/nm². The solid lines are the fits of the dimerization model to the data points prior to their binning. (A) CHO, HEK293T and A431 vesicles, produced with the DTT/formaldehyde method. (B) CHO vesicles, produced with the DTT/formaldehyde method and the chloride salt osmotic method. (C) A431 cells vesiculated using the three different methods.

A two-state model describing protein dimerization was used to fit the experimental dimeric fractions and calculate the dimerization constants. The predicted dimeric fractions were calculated as a function of the total concentration [T], based on the dimerization model given by equations (5) to (7), for any value of the dimerization constant K_D. Then, the value of K_D was optimized using a Matlab code such that the prediction gave the best fit to the single-vesicle dimeric fractions (prior to binning). The fits are shown in Figure 2 with the solid lines, allowing a direct comparison between the binned data and the fits. The optimal K_D values and the corresponding dimerization free energies (calculated using equation 9) are shown in Table 1. We see that the cell type and the vesiculation method have a very modest effect on GpA dimerization, not exceeding ~0.8 kcal/mole.

Table 1.

Dimerization constants and dimerization free energies measured for GpA in plasma membrane derived vesicles. By ANOVA, there are no significant differences between the measured free energies in the complete data set, and there are no significant pairwise differences, either.

Cell line	Vesiculation method	K (nm2/protein)	ΔG (kcal/mol)
CHO	DTT/formaldehyde	312	−3.4±0.2
CHO	chloride salt	526	−3.7±0.2
HEK293T	DTT/formaldehyde	312	−3.4±0.2
A431	DTT/formaldehyde	204	−3.2±0.2
A431	salt cocktail	769	−4.0±0.2
A431	chloride salt	625	−3.8±0.3

The six different free energies were compared using ANOVA as described in Materials and Methods. The ANOVA analysis showed that there were no significant differences between the measured free energies overall, and there were no significant pairwise differences, either. In addition, the ANOVA analysis of the separate data sets shown in Figures 2A, 2B, and 2C, shows no statistical significant differences, either. As GpA dimerization strength in A431 vesicles is similar to the strength in CHO and HEK vesicles, there is no evidence of promiscuous interactions between GpA and EGFR due to the presence of GxxxG motifs in both proteins. If a previously reported value for GpA dimerization in CHO vesicles produced via the formaldehyde/DTT method, −3.9 ± 0.2 kcal/mole (15), is included with the current data set, the ANOVA analysis of all of the free energy measurements shows no significant difference between this past measurement and the new data set.

Discussion

Plasma membrane derived vesicles give us the opportunity to study the interactions between complex membrane proteins. As the genes are introduced into mammalian cells, the recombinant membrane proteins are synthesized and folded correctly, and then glycosylated and trafficked to the plasma membrane. Thus, the interaction strength between native glycosylated MPs can be quantified in a native-like membrane without the need for their extraction, purification, and reconstitution. The lack of control over the composition of the vesicles, however, is a shortcoming of this technique. Also, we do not yet completely understand the apoptotic processes which are triggered by the buffer stress and result in vesiculation. The plasma membrane derived vesicles lack a transmembrane potential and a cytoskeleton, and are not a perfect mimic of the plasma membrane. Yet, plasma membrane derived vesicles are the only model system that has allowed us thus far to characterize the interaction strength between glycosylated membrane proteins in quantitative terms (16, 23).

To learn more about the utility and the limitations of the plasma membrane derived vesicles, here we characterized the interactions of GpA in the six different vesicle preparations. The goal of the study was to learn whether GpA interactions differ in different plasma derived vesicles, and whether a particular vesicle preparation should be preferred over others. We measured very similar GpA dimerization free energies in all vesicle preparations (Table 1). Indeed, the ANOVA analysis shows that the observed differences are not statistically significant. Thus, the different preparation methods yielded equivalent membrane model systems for studies of GpA interactions. Considering GpA as a general membrane protein model, our results suggest that any of the six vesicle preparations can be used to produce a relevant data set for membrane proteins.

Highlights.

• We study the utility of plasma membrane derived vesicles as a model membrane system.

• We measure GpA dimerization in vesicles from CHO, HEK 293T and A431 cells.

• We characterize the effect of the vesiculation methods on GpA dimerization.

• The vesicle preparation protocol has no significant effect on GpA dimerization.